SQL count consecutive days

Question

This is the SQL database data:

UserTable

UserName    | UserDate      | UserCode
-------------------------------------------
user1       | 08-31-2014    | 232
user1       | 09-01-2014    | 232
user1       | 09-02-2014    | 0
user1       | 09-03-2014    | 121
user1       | 09-08-2014    | 122
user1       | 09-09-2014    | 0
user1       | 09-10-2014    | 144
user1       | 09-11-2014    | 166
user2       | 09-01-2014    | 177
user2       | 09-04-2014    | 188
user2       | 09-05-2014    | 199
user2       | 09-06-2014    | 0
user2       | 09-07-2014    | 155

Should only count consecutive days (as Result) if [UserCode] is something else than zero. UserDate is between 09-01-2014 and 09-11-2014. Show result only if Result is 2 or more.

What I want to my sql query to return is:

UserName    | StartDate     | EndDate       | Result
----------------------------------------------------------
user1       | 09-01-2014    | 09-03-2014    | 2
user1       | 09-08-2014    | 09-11-2014    | 3
user2       | 09-04-2014    | 09-07-2014    | 3

Is this possible using only SQL query?

What version of mssql are you using? – Arion Sep 30 '14 at 09:26 — Arion, Sep 30 '14 at 09:26

score 39 · Accepted Answer · edited Feb 20 '20 at 02:27

This is a Gaps and Islands problem. The easiest way to solve this is using ROW_NUMBER() to identify the gaps in the sequence:

SELECT  UserName,
        UserDate,
        UserCode,
        GroupingSet = DATEADD(DAY, 
                            -ROW_NUMBER() OVER(PARTITION BY UserName 
                                                        ORDER BY UserDate), 
                            UserDate)
FROM    UserTable;

This gives:

UserName    | UserDate      | UserCode   | GroupingSet
------------+---------------+------------+-------------
user1       | 09-01-2014    | 1          | 08-31-2014    
user1       | 09-02-2014    | 0          | 08-31-2014    
user1       | 09-03-2014    | 1          | 08-31-2014    
user1       | 09-08-2014    | 1          | 09-04-2014    
user1       | 09-09-2014    | 0          | 09-04-2014    
user1       | 09-10-2014    | 1          | 09-04-2014    
user1       | 09-11-2014    | 1          | 09-04-2014    
user2       | 09-01-2014    | 1          | 08-31-2014    
user2       | 09-04-2014    | 1          | 09-02-2014    
user2       | 09-05-2014    | 1          | 09-02-2014    
user2       | 09-06-2014    | 0          | 09-02-2014    
user2       | 09-07-2014    | 1          | 09-02-2014

As you can see this gives a constant value in GroupingSet for consecutive rows. You can then group by this colum to get the summary you want:

WITH CTE AS
(   SELECT  UserName,
            UserDate,
            UserCode,
            GroupingSet = DATEADD(DAY, 
                                -ROW_NUMBER() OVER(PARTITION BY UserName 
                                                            ORDER BY UserDate), 
                                UserDate)
    FROM    UserTable
)
SELECT  UserName,
        StartDate = MIN(UserDate),
        EndDate = MAX(UserDate),
        Result = COUNT(NULLIF(UserCode, 0))
FROM    CTE
GROUP BY UserName, GroupingSet
HAVING COUNT(NULLIF(UserCode, 0)) > 1
ORDER BY UserName, StartDate;

Example on SQL Fiddle

Thanks for this but I edited my question. My mistake! I could not use sum because the UserCode is not always just 1 or 0. — majukivi, Sep 30 '14 at 10:17
OK, I have just changed it from `SUM` to `COUNT(NULLIF(UserCode, 0))`, so when `UserCode` is `0`, the `NULLIF` function will turn it into `NULL`, and it won't be included in the `COUNT` — GarethD, Sep 30 '14 at 10:26

TechDo · Answer 2 · 2014-09-30T10:27:18.787

0

Please try:

;with T1 as(
    select 
        *, 
        ROW_NUMBER() over (  order by UserName, UserDate) ID 
    from tbl
)
,T as (
    SELECT *, 1 CNT FROM T1 where ID=1
    union all
    SELECT b.*, (case when T.UserDate+1=b.UserDate and 
                           T.UserName=b.UserName then t.CNT 
                        else T.CNT+1 end)
    from T1 b INNER JOIN T on b.ID=T.ID+1
)
select distinct UserName, MIN(UserDate), max(UserDate)
,sum(case UserCode when 0 then 0 else 1 end) From T group by UserName, CNT
having COUNT(*)>1

SQL Fiddle Demo

edited Sep 30 '14 at 10:27

answered Sep 30 '14 at 09:54

TechDo

18,398
3
51
64

Thanks for this but I edited my question. My mistake! I could not use sum because the UserCode is not always just 1 or 0. – majukivi Sep 30 '14 at 10:17

SQL count consecutive days

2 Answers2

Linked

Related